Short Answer: The 10-80-10 rule is a practical way to divide work between people and AI. You invest the first 10% setting the direction, giving AI the right context and defining how the task should be done. AI handles the middle 80% of the repeatable execution, then the final 10% comes back to a human for judgement, refinement and approval.

In sales, that could mean AI researching a lead, organising information, drafting an email, preparing a proposal or reminding you to follow up. But decisions about fit, pricing, scope, relationships and what actually gets sent stay with you.

The percentages are not meant to be a rigid formula for every task. They are a useful design principle: use AI for the heavy lifting without automating away the parts of sales where human judgement matters most.

The Wrong Question Is "Can AI Do My Sales?"

AI can already handle far more of the sales process than most businesses are using it for. It can monitor inquiries, research prospects, organise CRM information, draft responses, prepare proposals, surface unanswered leads and help with onboarding administration.

But asking "Can AI do my sales?" is too broad.

A much better question is:

"Which parts of this sales process should AI handle, and which parts should remain human?"

That distinction matters because sales is not one task. It is a chain of very different activities, from capturing an inquiry and researching a lead through to pricing, follow-up, negotiation and onboarding.

Some of those tasks are repetitive, structured and ideal for AI. Others rely on discernment, emotional intelligence, trust and accountability.

The goal is not maximum automation. The goal is the right division of work.

The First 10%: Give AI Better Direction

The first 10% happens before AI starts doing anything.

This is where you define the task, outcome, context, boundaries and standards. Depending on the job, that could include an SOP, examples of good work, your products and services, FAQs, pricing rules, ideal client information, proposal templates or instructions about what AI must never decide on its own.

Think about delegating to a capable new team member. "Handle sales" would be a terrible brief. "Research this prospect, compare their likely needs against these three services, prepare a one-page brief and flag anything you're unsure about" gives them something they can actually execute.

AI works the same way.

The better the first 10%, the less intervention you should need during the middle 80%.

For example, if you want AI to help draft proposals, give it the proposal structure, relevant products and services, sales-call transcript, pricing guidance, examples and the reasoning behind important decisions. If you want it handling incoming inquiries, give it your target client information, FAQs, objections, tone and rules for when an inquiry needs to be escalated.

This is why businesses that simply switch on automation often end up disappointed. They try to automate execution before documenting how good work should happen.

The Middle 80%: Let AI Do the Heavy Lifting

Once the task is properly defined, this is where AI earns its place.

The middle 80% is the repetitive execution that consumes time but does not necessarily require your personal judgement at every step.

In a sales process, AI might help with:

• monitoring and categorising incoming inquiries

• gathering lead information

• researching a prospective client

• organising call notes or transcripts

• matching likely needs to relevant offers

• drafting personalised responses

• preparing proposal drafts

• surfacing relevant testimonials

• tracking follow-up dates

• reminding you about unanswered proposals

• preparing onboarding information

• organising data across connected systems

The key word is execution.

If you have briefed the task well, you should not need to manually guide AI through every tiny step. Otherwise, you have simply created a faster assistant that still requires constant supervision.

The opportunity with AI agents is to move beyond "give me the next paragraph" towards "here is the job, here are the resources and here is what good looks like - now complete the defined workflow."

If you're unsure which business tasks are ready for that level of delegation, the T.A.S.K framework for AI agents helps you assess whether a task is repeatable, properly resourced and suitable for AI before you start automating it.

The Final 10%: Keep the Decisions That Matter Human

This is the part businesses should be very careful not to remove simply because technology makes it possible.

The final 10% is where human judgement comes back in.

In sales, that could mean reviewing the final email, confirming the proposal scope, checking pricing, approving a discount, deciding whether a prospect is the right fit, reviewing contract terms or deciding whether another follow-up would be helpful or irritating.

AI can prepare those decisions. It should not automatically own every one of them.

A beautifully formatted proposal can still contain the wrong scope. A polished email can still sound inappropriate for that particular relationship. A confident AI recommendation can still be based on incomplete information.

Finished-looking is not the same as finished.

Human review adds something AI does not reliably reproduce: context that may never have been written down, relationship history, instinct, emotional intelligence and accountability for what happens next.

For client-facing work, a simple operating principle is useful:

AI drafts. Humans review. Humans publish.

What Does 10-80-10 Look Like in a Real Sales Process?

The model becomes easier to apply when you break sales into five stages:

Capture → Listen → Offer → Sale → Engage

At each stage, ask the same three questions:

1. What direction does the human need to provide?

2. What repetitive execution can AI take over?

3. What judgement must come back to a human before anything consequential happens?
   

Capture: Respond to the lead

Your 10% might include defining your ideal client, FAQs, common objections and rules for handling inquiries.

AI's 80% might monitor an inbox, categorise the inquiry, create a CRM record, gather relevant context and prepare a draft response.

Your final 10% is deciding whether the lead is actually a fit and reviewing the response before it goes.

Listen: Understand what they need

Your first 10% defines what information matters when assessing a potential client.

AI can then research the organisation, prepare a lead brief, identify likely needs and suggest questions for the conversation.

The human still has the call.

That matters because listening is not simply extracting information. A conversation includes tone, hesitation, priorities and signals that can change what you recommend.

Offer: Prepare the proposal

Your direction includes your proposal structure, products, services, pricing principles, examples and decision rules.

AI can combine that information with the research and call transcript to prepare a tailored proposal and associated email.

Your final 10% is where you confirm price, scope, discounts, commitments and whether the proposed solution is genuinely right for the client before anything is sent.

Sale: Follow up intelligently

You decide what an appropriate follow-up rhythm looks like and when a lead should no longer be actively pursued.

AI can track unanswered proposals, bring stalled conversations back to your attention and prepare contextual follow-ups based on what has already been discussed.

You decide whether the message goes, whether you should pick up the phone or whether it is time to stop.

Engage: Onboard the client

Once the client says yes, AI can take on a surprising amount of administration. It can help prepare project information, draft onboarding emails, organise client records and prepare documents from approved information.

But important details such as terms, price, scope, dates and contractual commitments still deserve human confirmation.

If AI Can Do It, Should You Automate It?

Not necessarily.

This is one of the most important distinctions in AI adoption.

Technical capability does not automatically equal good business design.

You may be able to automate an entire client conversation, but your clients might value speaking directly with you. You may be able to let AI send proposals automatically, but the financial and reputational consequences of an incorrect price might make that a poor choice.

Before automating any sales task, ask whether it is:

• repeatable enough to define clearly

• supported by reliable information

• easy to review before consequences occur

• recoverable if something goes wrong

• low enough risk for the permissions you are giving AI

If the answer is no, keep more of that task human.

The aim is not to prove how much AI can do. It is to build a sales process that works better because AI and humans are each doing the work they are best suited to.

Human Judgement Is Becoming More Valuable, Not Less

As AI gets better at execution, judgement becomes more important.

When creating a proposal manually took several hours, much of your attention was consumed by gathering information, formatting documents and assembling the first draft. If AI can take on most of that work, your time moves towards reviewing the recommendation itself.

Is this actually the right service?

Is the client a good fit?

Does the price make sense?

Is there something in the conversation that changes how we should approach this?

Should we send another follow-up or leave them alone?

These are not administrative questions. They are business decisions.

AI does not make your expertise less important. It gives you the opportunity to spend more of your time using it.

Don't Automate a Broken Sales Process

There is one catch.

AI cannot rescue a process you have never defined.

If proposals look completely different every time, nobody knows when leads should be followed up, pricing rules live entirely in someone's head and your sales information is scattered across five systems, automating the process may simply make the confusion happen faster.

Start with one repeatable task.

Document how you currently do it. Identify the inputs. Define what a successful output looks like. Decide which decisions need human approval.

Then automate the middle.

If you're ready to move from identifying the task to building the workflow, our guide to building your first AI agent walks through how to brief, test and deploy an agent while keeping a clear human review process in place.

The Best AI Sales System Should Still Feel Human

Customers should not have to experience your internal automation.

They should experience a business that responds quickly, remembers what they said, prepares properly, follows through and communicates with relevance.

That is the promise of good AI-powered sales.

The technology handles more of the invisible work underneath the relationship. The human has more capacity for the parts the customer actually values.

So instead of asking how much of your sales process you can automate, start with a better question:

What could AI take off my plate so I can spend more time on the decisions and relationships that actually need me?

That is what the 10-80-10 rule is designed to protect.

Ready to put AI to work without losing the human part?

You do not need to automate your entire sales process. Start with one repetitive task, give AI clear direction, let it handle more of the execution and keep your judgement at the final checkpoint.

That small shift can turn AI from another tool you have to manage into practical support for the work already happening in your business.

Our free AI workshop shows you how to identify the right tasks for AI and start building practical systems around them, without unnecessary complexity.

👉 Join our free AI workshop here.

Frequently Asked Questions

What is the 10-80-10 rule for AI?

The 10-80-10 rule is a practical framework for dividing work between humans and AI. The first 10% is human direction and context, the middle 80% is AI execution, and the final 10% returns to the human for judgement, refinement and approval. The percentages are a useful operating model rather than a rigid requirement for every task.

What parts of sales can AI automate?

AI can help with repetitive sales work such as lead research, inquiry categorisation, drafting responses, organising information, proposal preparation, follow-up reminders and onboarding administration. The exact level of automation should depend on the task, information available and consequences if something goes wrong.

What parts of sales should stay human?

Relationship conversations, fit decisions, sensitive negotiations, pricing exceptions, final proposal approval, contractual commitments and other decisions requiring judgement should generally retain clear human involvement.

Does human review defeat the purpose of automation?

No. Human review lets AI handle much of the repetitive preparation while you focus on the higher-value decision. Reviewing a completed draft is very different from manually creating everything from scratch.

How do I know which sales task to automate first?

Start with something repetitive, clearly defined and easy to review. Look for work that happens regularly, follows similar steps and already has enough information or examples for AI to understand what a good result looks like. Avoid starting with your highest-risk or most relationship-sensitive task.